I came to graduate school intent on becoming a mathematical psychologist and was going to work with Gordon Bower, one of the authors of a prominent textbook on mathematical psychology. On my first visit to his office he informed me that mathematical psychology was dead (a greatly exaggerated demise) and that I should go into artificial intelligence. The AI courses I took at Stanford presented an inspiring image of AI as pursuing the goal of a unified characterization of intelligence. While much of AI has abandoned this goal of unification, I imprinted on a similar but different goal of trying to achieve a more unified understanding of human cognition. My major accomplishment is what I have been able to contribute to this goal, currently realized in a cognitive architecture called ACT-R.

The first major step in this effort was the development with Gordon Bower of the HAM (Human Associative Memory) theory of human memory. We showed that many of the then fashionable effects in the study of human memory could be simulated by a computer system that just followed the links in memory. It did not take us long after completing the HAM theory, however, to recognize its major shortcoming. Just as Guthrie had criticized Tolman for leaving his rat buried in thought, we left the human buried in its memories with no course of action.

Allen Newell had been working on production systems to provide a principled theory of how human knowledge resulted in action. A production system is composed of many statements similar to: IF one's memory is in state x, THEN do y. Such a system essentially organizes set of cognitive reflexes into coherent cognitive behavior. The first published version of an ACT (Adaptive Control of Thought) theory was essentially a production system that responded to the declarative memories of HAM. Newell's work drew me to Carnegie Mellon, where I have spent over half my adult life. From Newell I learned about his conception of a cognitive architecture and realized that this was what I had been striving for with ACT. Borrowing heavily from the ideas in my environment, I developed a version of the ACT* theory, which I thought was “the final major reformulation within the ACT framework.” This was an incorrect projection to match Gordon Bower's from fifteen years earlier.